Granular fingering as a mechanism for ridge formation in debris deposits: Laboratory experiments and implications for Tutupaca volcano, P. Valderrama, Olivier Roche, P. Samaniego, Benjamin van Wyk de Vries, G. Araujo

To cite this version:

P. Valderrama, Olivier Roche, P. Samaniego, Benjamin van Wyk de Vries, G. Araujo. Granular fingering as a mechanism for ridge formation in debris avalanche deposits: Laboratory experiments and implications for Tutupaca volcano, Peru. Journal of Volcanology and Geothermal Research, Elsevier, 2018, 349, pp.409-418. ￿10.1016/j.jvolgeores.2017.12.004￿. ￿hal-01677525￿

HAL Id: hal-01677525 https://hal.uca.fr/hal-01677525 Submitted on 2 Oct 2018

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Granular fingering as a mechanism for ridge formation in debris

avalanche deposits: Laboratory experiments and implications for

Tutupaca volcano, Peru

a,b b,⁎ b b a P. Valderrama , O. Roche , P. Samaniego , B. van Wyk des Vries , G. Araujo a Observatorio Vulcanológico del INGEMMET, Av. Canada 1470, San Borja, Lima, Peru b Université Clermont Auvergne, CNRS, IRD, OPGC, Laboratoire Magmas et Volcans, F-63000 Clermont-Ferrand, France

The origin of subparallel, regularly-spaced longitudinal ridges often observed at the surface of volcanic and other rock avalanche deposits remains unclear. We addressed this issue through analogue laboratory experiments on flows of bi-disperse granular mixtures, because this type of flow is known to exhibit granular fingering that causes elongated structures resembling the ridges observed in nature. We considered four different mixtures of fine (300–400 μm) glass beads and coarse (600–710 μmto900–1000 μm) angular crushed fruit stones, with particle size ratios of 1.9–2.7 and mass fractions of the coarse component of 5–50 wt%. The coarse particles seg- regated at the flow surface and accumulated at the front where flow instabilities with a well-defined wavelength grew. These formed granular fingers made of coarse-rich static margins delimiting fines-rich central channels. Coalescence of adjacent finger margins created regular spaced longitudinal ridges, which became topographic highs as finger channels drained at final emplacement stages. Three distinct deposit morphologies were ob- served: 1) Joined fingers with ridges were formed at low (≤1.9) size ratio and moderate (10–20 wt%) coarse frac- tion whereas 2) separate fingers or 3) poorly developed fingers, forming series of frontal lobes, were created at larger size ratios and/or higher coarse contents. Similar ridges and lobes are observed at the debris avalanche de- posits of Tutupaca volcano, Peru, suggesting that the processes operating in the experiments can also occur in na- ture. This implies that volcanic (and non-volcanic) debris can behave as granular flows, which has important implications for interpretation of deposits and for modeling. Such behaviour may be acquired as the collapsing material disaggregates and forms a granular mixture composed by a right grain size distribution in which particle segregation can occur. Limited fragmentation and block sliding, or grain size distributions inappro- priate for promoting granular fingering can explain why ridges are absent in many deposits.

Keywords: Debris avalanche Granular fingering Ridge Granular flow Analogue experiments

1. Introduction of meters. Examples have been described in volcanic debris avalanche deposits (Fig. 1) such as at (Belousov et al., 1999)and Debris avalanches and landslides occur in many volcanic and non- (van Wyk de Vries et al., 2001), and in landslides in non- volcanic contexts and represent major geological hazards (van Wyk volcanic environments such as the Sherman rock-avalanche, de Vries and Davies, 2015). The kinematics and dynamics of these grav- Alaska (Dufresne and Davies, 2009) or the Tschirgant rock-avalanche, itational mass flows can be studied through analysis of the architecture Austria (Dufresne et al., 2016). Similar features have also been described and the surface structures of their deposits. The geological literature re- in extraterrestrial contexts (Luchitta, 1978). ports various surface structures, including hummocks (Clavero et al., Samaniego et al. (2015) and Valderrama et al. (2016) have reported 2002; Paguican et al., 2012) and elongated structures termed either fur- very well preserved ridges in the last debris avalanche deposit (218 ± rows (Belousov et al., 1999), flow bands (Dufresne and Davies, 2009)or 14 aBP) of Tutupaca volcano in southern Peru (Fig. 1). Valderrama ridges (or longitudinal ridges, Samaniego et al., 2015; Valderrama et al., et al. (2016) concluded that these ridges may give fundamental insights 2016). The elongated structures, hereafter designated as ridges, are into the dynamics of the debris avalanche, as they could be the conse- strictly linear to slightly curved, often form swarms of subparallel and quence of granular fingering, a physical process observed in poly- regularly-spaced lineations, and have typical lengths of several hundred disperse granular flows (Pouliquen et al., 1997; Pouliquen and Vallance, 1999; Malloggi et al., 2006; Woodhouse et al., 2012; Gray ⁎ Corresponding author. et al., 2015). The present work is aimed at further investigating granular E-mail address: [email protected] (O. Roche). fingering in order to discuss the mechanisms of debris avalanches. It is

1 Fig. 1. Longitudinal ridges at the surface of the debris avalanche deposits. (a) Shiveluch volcano (Belousov et al., 1999), (b) Tutupaca volcano (Google Earth image, Valderrama et al., 2016).

worth noting that granular fingering has received little attention since 2. Fingering: a fundamental process in granular flows its discovery by Pouliquen et al. (1997) and its detailed mechanisms have yet to be fully investigated. For this reason we made new experi- Granular fingering was first reported by Pouliquen et al. (1997) from ments on granular flows using bi-disperse mixtures as in previous sim- a series of laboratory experiments, and it was further investigated by ilar studies but considering systematically different grain size ratios and Pouliquen and Vallance (1999), Malloggi et al. (2006), Woodhouse concentrations of the granular components. We report the different et al. (2012) and Gray et al. (2015). In their seminal studies Pouliquen stages of granular segregation and fingering and discuss how the grain et al. (1997) and Pouliquen and Vallance (1999) carried out experi- size ratio influences the shape and the surface structures of the flow de- ments on bi-disperse granular flows on rough inclines (Fig. 2). They posits. Finally, we show that the structures in experiments share many used mixtures of small sub-spherical glass beads of diameter of 500 similarities with those described at the Tutupaca volcano debris ava- μm and of large particles made of angular crushed fruit stones of size lanche, which permits us to interpret the emplacement mechanisms of 570 μmor1200μm, with volume concentrations of the coarse com- of this volcanic mass flow. ponent of 5–8%. Flows were generated in a channel, 1.5–2 m-long and

Fig. 2. Principles of granular fingering. (a) Formation of granular fingers at successive times (i–iii) in a mixture of 95 vol% of 500 μm diameter beads and 5% vol% of 570 μm diameter angular particles (black) flowing on a rough substrate inclined at 24.5° (in Pouliquen and Vallance, 1999). (b) Trajectories of the coarse particles on top (black arrows) or bottom (white arrows) of the granular flow (in Pouliquen et al., 1997). (c) Sketch of the recirculation of the coarse particles at the flow front (from Pouliquen et al., 1997).

2 0.7 m-wide, whose rigid substrate was made rough by gluing the same onto an adhesive sheet that laid on the channel base. As some particles particles as the fine granular flow component. The bi-disperse mixtures could be entrained by the flows, the base was replaced every ten exper- were released from a reservoir with constant gate opening of iments to ensure a consistent roughness. The experiments were filmed 0.4–0.7 cm, hence generating flows of nearly constant thickness down with a high resolution (1920 × 1080 pixels) video camera in order to in- the incline. Granular fingering arose as a consequence of particle segre- vestigate the flow kinematics and the segregation processes. Transver- gation. The larger particles segregated at the surface of the flows be- sal black marks were placed beneath the non-opaque rough base cause of the combined effects of percolation of the small beads every 5 cm to allow for measurement of the flow front propagation. through the shearing granular network and of squeeze expulsion of Experiments were done with four bi-disperse mixtures of sub- the large components. Once at the flow surface, these large particles spherical glass beads with a grain size of 300–400 μm (same as for the travelled faster than the rest of the flowing mass, because the flow ve- rough channel base) and of coarse angular particles (crushed fruit locity across the flow depth increased upwards, and as a consequence stones, cf. Pouliquen et al., 1997; Pouliquen and Vallance, 1999) with they concentrated at the flow front. The large particles were deflected grain size ranges from 600 to 710 μmupto900–1000 μm(Table 1). along the steepest surface slope at the flow front and accumulated to The mass fraction of the coarse component in the mixtures varied form local instabilities characterized by emerging frontal lobes with from 5 to 50%. Before each experiment, the bi-disperse material was static lateral margins immediately behind. Large friction associated to stirred by hand gently and thoroughly to generate mixtures as homoge- the sliding of these irregular-shaped particles helped amplifying the neous as possible. The fine and coarse components of the mixtures rep- frontal instabilities, which progressively acquired the shape of granular resented respectively the matrix and the largest blocks of natural fingers of nearly constant width and with well-defined static coarse- materials. grained margins that merged to form well-defined linear structures. Segregation and granular fingering were favoured by recirculation of 3.2. Photogrammetry of the deposits the coarse grains at the flow front as particles that reached the flow base were reinjected upwards to the free surface where segregation After each experiment at least 14 photographs were taken with the acted again (Fig. 2). This recirculation process was later investigated in high resolution camera from different angles for photogrammetry in detail in large-scale experiments by Johnson et al. (2012) and theoreti- order to obtain digital elevation models (DEM) of the flow deposits. A cally by Gray and Kokelaar (2010).Granularfingering was also ad- white halogen light was used to make stand out the detailed shapes dressed theoretically by Woodhouse et al. (2012) and Gray et al. and structures of the deposits. Photogrammetry was processed using (2015). Pouliquen and Vallance (1999) and Malloggi et al. (2006) re- the software Agisoft Photoscan Professional Edition version 1.1.6 for ported that granular fingering can occur as well in subaqueous flows. Mac (see Smith et al., 2016, for details on the method). In order to im- A key result of the studies is that the merged coarse-grained margins prove the DEM resolution, we created a reference system with a mini- of granular fingers are longitudinal structures similar to the ridges ob- mum unit size of 0.05 mm, which was calibrated using the served at the surface of many debris avalanche deposits. These studies, coordinates of 7 fixed points previously marked on the experimental however, focused on the morphology of the flowing granular masses device. Data processing permitted us to obtain DEMs of the experimen- and did not consider their deposits. Furthermore, they did not investi- tal deposits with a resolution of 0.5 mm (Fig. 4). The DEMs were then gate granular fingering systematically as a function of the size ratio of used to retrieve cross-sections of the deposits, which were compared the granular components and they considered only very low volume to measurements made in situ with a ruler. fractions of the coarse grains. In this context, the aim of our experimen- tal study is to further investigate granular fingering by taking into ac- 4. Results count wider ranges of grain size ratio and of proportion of the components of bi-disperse mixtures. Also, we investigated the flow ki- In this section we first present preliminary experiments that permit- nematics as well as the morphology of the deposits, which can be com- ted us to define the parameters we chose to investigate granular pared to natural cases. fingering in detail. Then, we report the flow kinematic data as well as the morphological characteristics of the deposits, including the final 3. Experimental procedure length, the thickness and the width of the fingers, and the distance be- tween the axis of the fingers. 3.1. Experimental device and particles 4.1. Preliminary tests The experimental device consisted of a 1.45 × 0.45 m inclined chan- nel connected to a reservoir from which granular mixtures were re- We tested different mixtures with various grain sizes and propor- leased to generate flows (Fig. 3). The reservoir was equipped with a tions of particles. We first used mixtures of the glass beads (300–400 double gate system to ensure a constant mixture outflow rate. The aper- μm) and the smaller coarse particles (600–710%), the latter at concen- ture of the back door was set to 1.5 cm for all experiments while the trations of 5 to 50 wt% (Table 1). On the basis of this first series of exper- front door was entirely removed to release the particles. The channel iments, we concluded that the flows experienced well-defined granular was made rough by gluing 300–400 μm subspherical glass particles segregation and fingering for coarse particle concentrations of 5 to 20 wt%. Then, experiments were made with larger coarse particles sizes at concentrations of 5 to 20 wt% (Table 1). Experiments with a given mixture were repeated at least three times and showed good reproducibility. The inclination of the channel and the volume of material released from the reservoir were set according to the experiments that generated the most elongated deposits but shorter than the channel length. We concluded that the most appropriate slope angle and material mass were 31° and 1 kg, respectively. At lower angles the flows stopped too close to the gate while at higher angles they left the channel. The mass (and so the volume) of material controlled fundamentally the length of the deposits. Videos of the experiments revealed that the flow kine- Fig. 3. Schematic view of the experimental device. matics was characterized by three stages of emplacement: particle

3 Table 1 Granular bi-disperse mixtures used in experiments. Grain size ranges were obtained through particle sieving.

Fine particles (glass beads) Coarse particles (crushed fruit stones) Mean size ratio Mass fraction of large particles (%)

Mixture 1 300–400 μm 600–710 μm 1.9 5, 10, 15, 20, 30, 40, 50 Mixture 2 300–400 μm 710–800 μm 2.1 5, 10, 15, 20 Mixture 3 300–400 μm 800–900 μm 2.4 5, 10, 15, 20 Mixture 4 300–400 μm 900–1000 μm 2.7 5, 10, 15, 20

segregation, then accumulation of the coarse particles at flow front, and short, typically 5–15 cm, and the flows propagated at relatively slow ve- finally granular fingering (Fig. 5). This is described in detail below (see locities of 5–15 cm/s. The complementary experiments with ≥30 wt% of Fig. 6). 600–710 μm coarse particles, however, revealed that the distance trav- elled was up to ~70 cm at concentration of 50 wt%. 4.2. Flow kinematics 4.2.3. Stage 3: fingering 4.2.1. Stage 1: segregation The final stage of propagation was characterized by granular Particle segregation occurred as soon as the granular mixture prop- fingering and a significant decrease of the velocity of the flowing mass agated downslope after gate opening (Fig. 6). Most of the small particles (Figs. 6 and 7). The fingers formed as a consequence of particle segrega- percolated downwards while the coarse particles moved to the top of tion, accumulation of coarse particles at the flow front and formation of the flow and concentrated at the front. This stage was characterized local instabilities, which grew to form fines-rich channels bordered by by the highest flow front velocity, which was about 40 cm/s and was in- coarse-rich static margins (cf. Pouliquen et al., 1997). Accumulation of dependent on both the coarse particle size and concentration (Fig. 7). It the coarse particles at the flow front, which began during stage 2 and appeared that at this stage, the front velocity was controlled by the decreased the velocity of the moving mass, further operated during slope angle. The distance travelled by the flows, however, tended to de- this last stage and caused even lower propagation velocities of about crease with the coarse particles content (Fig. 7). 2–3 cm/s until the granular mass halted (Fig. 7). Note that some low am- plitude surface waves that arose in the proximal area provided material 4.2.2. Stage 2: accumulation of coarse particles at flow front that concentrated further downstream in the channels (see Fig. 6 at t During this stage most of the coarse particles accumulated at the N 6.16 s). The fingers delimited by the stable margins had a constant flow front (Fig. 6). This caused instabilities (cf. Pouliquen et al., 1997) width, and the flowing granular mass mainly composed of small parti- and the flow front slowed down (or even stopped for a short duration cles in the central channel eventually drained once material supply in some experiments) while the rest of mass upstream propagated at from the reservoir was no longer available. The inner part of the static the initial velocity acquired at Stage 1. In consequence, the front was margins could be eroded by the flow in the central channels. overtaken by the material behind, which caused a surface wave which An important result of our experiments is that the fingers were ei- favoured flow propagation downslope. The front instabilities then ther merged or separated depending on the nature of the granular mix- grew to form emerging fingers. After a couple of seconds, about 80% of tures (Fig. 8). Well-defined long fingers were merged, and their joint the coarse particles had accumulated at the front (Fig. 6, t =2.06s). margins then formed longitudinal ridges, when the grain size difference Detailed information on the times and distances at which this sec- between the two components was small (i.e. with 600–710 μm coarse ond stage began and ended for the different mixtures is presented in particles) and at moderate coarse concentrations of 10–20 wt% (lower Figs. 6 and 7. For mixtures with ≤20 wt% of coarse particles stage 2 concentrations of 5 wt% did not promote enough segregation for began at distances of 60–80 cm from the reservoir and at times of allowing emergence of fingering). With these conditions the mean dis- 2–3 s after release, except for mixtures with 900–1000 μmcoarseparti- tance travelled by the fingers was typically 20–30 cm (Fig. 6). Note that cles for which these distances and times were rather 75–95 cm and at concentrations of 20 wt% the range of travel distance was fairly large 3–3.5 s, respectively. During this stage the distance travelled was fairly and that some fingers propagated out of the channel (Fig. 8). In contrast,

Fig. 4. Example of a flow deposit. (a) Photograph showing the small glass beads (white) and the segregated coarse particles (brown). (b) Digital elevation model. (c) Contour lines parallel to the inclined plane. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

4 topographic highs as the granular material in the channel drained at late stages. In case of joined fingers the levees either merged completely or incompletely, hence leading to one large or two small local topo- graphic highs, and they thus formed the longitudinal ridges. When the local instabilities did not lead to fully developed fingers the gran- ular material accumulated at the front and formed series of merged, and sometimes superposed lobes thicker than the rest of the deposit upstream. The detailed morphological data of the fingers are given in Fig. 10. Note that we measured all fingers even when these were poorly devel- oped and rather resembled lobes (see Fig. 8). We recall that the percent- ages we give below are those of the coarse particles concentration in the mixtures. The joint fingers were the longest for mixture 1 with the smallest (600–710 μm) coarse particles. Their mean length increased from 35 cm at 5 wt% up to 63 cm at 20 wt%, and complementary exper- iments at higher concentrations revealed lengths of ~50–70 cm. Notice that the ranges of length were quite large, typically ~30–50 cm. For other mixtures the joint fingers were significantly shorter as their mean length was ~10–30 cm. For mixtures 2 and 3, however, one or two long (up to ~50 cm) fingers commonly formed, causing length ranges as large as for mixture 1. The separate fingers formed in mixtures 3 (at 10 wt%) and 4 (at fi fl 20 wt%) had signi cantly longer mean lengths of 37 cm and 39 cm, re- Fig. 5. Contours of the ow front at end of the three stages of emplacement. Stage 1: fi – segregation of the particles. Stage 2: accumulation of the coarse particles at flow front. spectively. The mean thickness of the ngers was ~2 6mmandvaried Stage 3: granular fingering. very little in a given experiment except at very high coarse particle con- tent of 50 wt% in mixture 1. It increases clearly with the coarse concen- tration in mixture 1 at ≥15 wt% and in mixture 2. The long separate in other mixtures the fingers were either separated and/or poorly devel- fingers in mixture 4 at 20 wt% were thinner (~1.5 mm) than in other oped and then rather formed series of frontal lobes. In particular, well- cases. The mean width of the fingers was 3.5–5.5 cm for mixtures at defined separated fingers formed in mixtures 3 and 4 with coarse com- 5–20 wt% (with ranges of values of ~2–3 cm). It increased fairly clearly ponents of 800–900 μm at 10 wt% and of 900–1000 μmat20wt% with the coarse particles content and was about the same for the differ- (Fig. 8). ent mixtures at same concentrations. Complementary experiments with mixture 1 showed that the width increased up to ~9 cm at 4.3. Morphology of the deposits 50 wt%. The widths of the separate fingers were in the trends defined by those of the joint fingers. Finally, the distance between the axis of Photogrammetric images of the three types of deposit morphology the joint fingers was in general close to the width, as imposed by the are shown in Fig. 9. They reveal that the fingers consisted of a frontal geometrical configuration. For the separate fingers, however, this dis- lobe and of a central channel bordered by lateral levees, which became tance was longer that the finger width.

Fig. 6. Snapshots showing the stages of flow propagation for a mixture of 300–400 μm glass beads and 15 wt% of 600–710 μm coarse particles. Final deposit at t =10.32s.

5 Fig. 7. Kinematic data of the flows of different granular mixtures. The grain size and the proportion of the coarse component are given. The symbols indicate the end of stages 1, 2 and 3. Note that some flow travelled out the channel (runout distance N 140 cm).

5. Discussion Note that the separated fingers are similar to the distal deposits of or debris flows (Jessop et al., 2012; Kokelaar et al., 2014), 5.1. Granular fingers and ridges in experiments which have well-defined levees bordering a central channel as well as a terminal frontal snout. Our experiments showed that longitudinal ridges were formed through coalescence of margins of adjacent fingers (Fig. 8). The ridges 5.2. Comparison with a natural case: The Tutupaca debris avalanche were well-developed when the fingers were long and the central chan- deposits nels had drained sufficiently at final stages of emplacement to cause a notable difference in height with the ridges. Merging of margins could We now compare our experimental results with observations made be incomplete and then led to ridges with two peak heights (Fig. 9). on natural debris avalanche deposits. We stress that the experiments The most favourable conditions for the formation of ridges in experi- allowed us to investigate the main factors controlling granular fingering ments were met when the grain size ratio between the two components in a configuration that represented an ideal simple case compared to of the granular mixture was ≤1.9 (i.e. mixture 1, see Table 1) and the natural systems. First, a high constant slope angle of 31° was required concentration of the large particles was 10–20 wt%. At lower concentra- in the models to generate slow granular flows from a steady release of tions there were not enough coarse particles to promote efficient segre- material initially at rest, whereas in nature high flow velocity and inertia gation and fingering. At higher concentrations, however, too many allow for propagation at much lower slopes angles. Also, topographic ir- coarse particles accumulated at the flow front, which inhibited fingering regularities such as break in slopes occurring frequently in nature may and could even stop motion, and in this case the front of the deposit alter significantly the flow dynamics (see Sulpizio et al., 2016). Second, consisted of series of more or less defined lobes (as observed for mix- granular fingering in poly-disperse natural materials could behave dif- tures with higher grain size ratios). ferently to the experimental bi-dispersed mixtures. Hence, although Series of separated fingersformedinmixtureswithgrainsizeratios the general trends of development of granular fingering observed in ≥2.1. Our results suggest that these structures could arise at increasing the experiments may be extrapolated to natural systems, we stress coarse particles content when the size ratio increased, i.e. in mixture 3 that only qualitative implications of the experimental results can be (size ratio of 2.4) at 10 wt% and in mixture 4 (size ratio of 2.7) at discussed at this stage. In other words, the critical grain size ratios and 20 wt% (see Fig. 8). It appears that the formation of separated fingers re- coarse particle contents that promoted granular fingering in the exper- quired an optimal range of concentration of large particles for a given iments might have different values in nature. grain size ratio, similarly to the merged fingers, otherwise segregation We consider the Tutupaca volcano located in southern Peru, which was not sufficient (at low concentration) or the high amount of large we investigated recently (Samaniego et al., 2015; Valderrama et al., particles at the flow front inhibited motion (at high concentration). 2016). The last historical (218 ± 14 aBP) eruption of Tutupaca generat- A fundamental difference between the separated and the merged ed a large sector collapse that triggered a debris avalanche and an asso- fingers is the wavelength of the instabilities that led to the emergence ciated pyroclastic eruption (Samaniego et al., 2015). The debris of individual fingers at the flow front. We highlight that this wavelength avalanche deposits are characterized by two distinct units: Unit 1 con- cannot be predicted by the model of Pouliquen and Vallance (1999) and taining a large amount of hydrothermally altered material that mostly that Woodhouse et al. (2012) argued that the wavelength observed in belongs to an older basal edifice, and Unit 2 consisting of fresh dacitic their numerical simulations is dependent on the number of grid points rocks from the youngest dome complex. Interestingly, the debris ava- used. Nevertheless, our data on the width of the fingers and on the dis- lanche deposits have different surface morphologies, including longitu- tance between their axes (section 4.3) suggest that this wavelength dinal ridges that border depressions and distal surface frontal lobes that may increase with the size ratio. are present in all but the proximal areas covered by hummocks (Fig. 11).

6 Fig. 8. Deposits in experiments with different granular mixtures. The rectangle indicates the most favourable conditions for the formation of merged fingers that generate ridges. The dashed rectangles point out well-developed separated fingers.

The analysis made by Valderrama et al. (2016) on N300 ridges revealed channels of the fingers drained at final stages of emplacement to form that these are 20–500-m long (mean of ~100 m) and 10–30-m wide, the depressions. Valderrama et al. (2016) concluded that these parts and their top is 1–5 m above the depressions. The distance between of the debris avalanche behaved like a granular flow, while the hum- the top of the ridges is 10–60 m (mean of ~30 m) and it increases mocky parts slid en-masse. Our new experimental data further support with the travel distance as the ridges fan slightly outward. A key obser- this conclusion and also give more insights into the granular behaviour vation is that the ridges have coarser cores and finer troughs, which sug- of the Tutupaca debris avalanche. gests grain size segregation during emplacement of the debris Direct comparison between the Tutupaca debris avalanche and the avalanche. Considering these structural and granulometric data experiments is not straightforward owing to the differences between Valderrama et al. (2016) argued that the ridges were formed through the two systems mentioned above. Our experimental results, however, granular fingering and resulted from merging of lateral margins of fin- suggest that polydispersity is not required to promote granular gers, and they became morphologically distinct when the central fingering and ridges, and that the main control factor is rather the

7 Fig. 9. Morphology of the experimental deposits obtained from photogrammetry. The three types of morphologies are shown with corresponding cross-sections. (a) Merged fingers (mixture 1, 600–710 μm at 15 wt%). (b) Separated fingers (mixture 4, 900–1000 μm at 20%). (c) Frontal lobes (mixture 2, 710–800 μm at 15%). Notation: levees (Le), ridges (Ri) and lobes (Lo). presence of large particles at optimal concentrations (notice that parti- segregation: joint fingers, separated fingers and frontal lobes, the latter cle angularity has only a second-order effect on fingering, cf. Pouliquen being poorly developed fingers. Separated fingers were analogues of et al., 1997).ThezoneAinFig. 11 shows a 1 × 2 km area of the Tutupaca pumice flow deposits with lateral levees bordering a central, less elevat- debris avalanche deposits with distinct subparallel ridges, at a distance ed channel. Joint fingers were created at low size ratio (≤1.9) and mod- of 4–5 km north of the amphitheatre. These ridges resemble those erate coarse particle content (10–20 wt%) and they led to the formation formed in experiments where coarse-rich margins of granular fingers of longitudinal ridges, which are the core theme of our study (note that coalesced. In nature, the varying underlying topography of the substrate joint fingers and ridges in nature could occur at size ratios and coarse on which the avalanche propagated as well as the possibility of the particle contents different than in experiments). In contrast, separated granular mass to spread radially led to non rectilinear ridges, in contrast fingers and lobes formed at higher size ratios and/or coarse particle con- to the simple configuration of the experiments. The distal zone B in tents. The factors controlling the wavelength of the flow front instabil- Fig. 11 shows another area at about 6 km north-west of the ities, which sets the width of the fingers and hence the distance amphitheatre, with a series of structures that can be interpreted as su- between the ridges, remain an open issue. Our results, however, suggest perposed frozen flow pulses that form frontal lobes. These structures that this wavelength may increase with the coarse particle content. are very similar to those in our experiments where granular fingering Though the experiments involved bi-disperse granular mixtures they could not develop because of accumulation of coarse particles at flow were able to reproduce deposit structures formed from poly-disperse front. The concentration of large blocks at front of the Tutupaca debris flows in nature, which suggests that large particles at optimal content avalanche may have caused the lobes observed in the field. is the key parameter for promoting fingering. This experimental study shows that ridges can form as a conse- 6. Conclusion quence of granular fingering, which itself is caused primarily by particle size segregation. The ridges arise because of coalescence of coarse-rich Our experiments, which involved larger ranges of size ratio and of lateral margins of joint fingers of nearly constant width and which coarse particles content compared to earlier studies, were used to ex- drain at late emplacement stages. Therefore ridges are well-defined to- plore granular fingering. They revealed three distinct morphologies of pographic highs spaced regularly. The main implication of the presence deposits of flows of bi-disperse mixtures that experienced particle size of ridges in deposits is that the parent debris avalanches behaved as

8 Fig. 10. Morphological characteristics of the granular fingers as a function of the coarse particles concentration, with mean values (squares) and ranges of data. The length and the thickness are that of the coarse-rich lateral levees. The size of the coarse particles in the mixtures is indicated, and open symbols stand for separate fingers. Note that no data are reported for the mixture with 5 wt% of 900–1000 μm coarse particles because no measurable structures were formed.

granular flows, which adds to a commonly accepted model of emplace- over flow and of the lack of material disaggregation, or grain size distri- ment by sliding of a coherent mass (see Shea and van Wyk de Vries, butions not suitable for granular fingering. 2008, and references therein). The Tutupaca debris avalanche deposits suggest that both the granular flow and block sliding mechanisms can Acknowledgment coexist during a given collapse event and lead to distinct surface mor- phologies. In this context the granular flow mechanism, favoured by This work is part of a collaborative programme between the high fragmentation of the collapsing material, may become dominant Peruvian Instituto Geológico, Minero y Metalúrgico (INGEMMET) and with the travel distance and therefore could promote granular flow in- the French Institut de Recherche pour le Développement (IRD). It was dicated by increasing amounts of ridges. The absence of ridges in partially founded by IRD through a PhD grant awarded to P. Valderrama. some deposits may be explained either by the dominance of sliding This research was also supported by the French Government Laboratory

9 Fig. 11. Map of Tutupaca volcano showing the two units of the 218 ± 14 aBP debris avalanche deposit (DAD) and their longitudinal ridges (from Valderrama et al., 2016), Google Earth images of zones A and B, and comparison with experimental results. A) Subparallel ridges of typical length of 100–500 m and interdistance of 10–30 m, and experimental deposit of the mixture with 15 wt% of 600–710 μm coarse particles showing ridges formed from merged margins of granular fingers. B) Frontal lobes in distal area, and experimental deposit of the mixture with 15 wt% of 710–800 μm coarse particles showing frontal lobes. of Excellence initiative no ANR-10-LABX-0006, the Région Auvergne Paguican, E.M.R., van Wyk de Vries, B., Lagmay, M.F., 2012. Hummocks: how they form and how they evolve in rockslide-debris avalanches. Landslides 11:67–80. https:// and the European Regional Development Fund. This is Laboratory of Ex- doi.org/10.1007/s10346-012-0368-y. cellence ClerVolc contribution no 278. We thank Domenico Doronzo for Pouliquen, O., Vallance, J.W., 1999. Segregation induced instabilities of granular fronts. useful comments. Chaos 9:621–630. https://doi.org/10.1063/1.166435. Pouliquen, O., Delour, J., Savage, S.B., 1997. Fingering in granular flows. Nature 386, 816–817. References Samaniego, P., Valderrama, P., Mariño, J., van Wyk de Vries, B., Roche, O., Manrique, N., Chédeville, C., Liorzou, C., Fidel, L., Malnatti, J., 2015. The historical (218 ± 14 aBP) ex- Belousov, A., Belousova, M., Voight, B., 1999. Multiple edifice failures, debris avalanches plosive eruption of Tutupaca volcano (Southern Peru). Bull. Volcanol. 77:51. https:// and associated eruptions in the history of Shiveluch volcano, Kamchatka, doi.org/10.1007/s00445-015-0937-8. . Bull. Volcanol. 61:324–342. https://doi.org/10.1007/s004450050300. Shea, T., van Wyk de Vries, B., 2008. Structural analysis and analogue modeling of the ki- Clavero, J.E., Sparks, R.S.J., Huppert, H.E., Dade, W.B., 2002. Geological constraints on the nematics and dynamics of rockslide avalanches. Geosphere 4:657–686. https:// emplacement mechanism of the debris avalanche, northern Chile. Bull. doi.org/10.1130/GES00131.00131. Volcanol. 64, 40–54. Smith, M.W., Carrivick, J.L., Quincey, D.J., 2016. Structure from motion photogrammetry in Dufresne, A., Davies, T.R., 2009. Longitudinal ridges in mass movement deposits. Geomor- physical geography. Prog. Phys. Geogr. 40:247–275. https://doi.org/10.1177/ phology 105:171–181. https://doi.org/10.1016/j.geomorph.2008.09.009. 0309133315615805. Dufresne, A., Bosmeier, A., Prager, C., 2016. Sedimentology of rock avalanche deposits - Sulpizio, R., Castioni, D., Rodriguez-Sedano, L.A., Sarocchi, D., Lucchi, F., 2016. The influ- case study and review. Earth Sci. Rev. 163, 234–259. ence of slope-angle ratio on the dynamics of granular flows: insights from laboratory Gray, J.M.N.T., Kokelaar, B.P., 2010. Large particle segregation, transport and accumulation experiments. Bull. Volcanol. 78:77. https://doi.org/10.1007/s00445-00016-01069- in granular free-surface flows. J. Fluid Mech. 652:105–137. https://doi.org/10.1017/ 00445. S002211201000011X. Valderrama, P., Roche, O., Samaniego, P., van Wyk de Vries, B., Bernard, K., Mariño, J., 2016. Gray, J.M.N.T., Gajjara, P., Kokelaar, P., 2015. Particle-size segregation in dense granular Dynamic implications of ridges on a debris avalanche deposit at Tutupaca volcano avalanches. C.R. Phys. 16:73–85. https://doi.org/10.1016/j.crhy.2015.1001.1004. (southern Peru). Bull. Volcanol. 78:14. https://doi.org/10.1007/s00445-016-1011-x. Jessop, D.E., Kelfoun, K., Labazuy, P., Mangeney, A., Roche, O., Tilliere, J.-L., Trouillete, M., Woodhouse, M.J., Thornton, A.R., Johnson, C.G., Kokelaar, B.P., Gray, J.M.N.T., 2012. Thibault, G., 2012. LiDAR derived morphology of the 1993 pyroclastic flow de- Segregation-induced fingering instabilities in granular free-surface flows. J. Fluid posits, and implication for flow dynamics and rheology. J. Volcanol. Geotherm. Res. Mech. 709, 543–580. 245–246:81–97. https://doi.org/10.1016/j.jvolgeores.2012.06.030. van Wyk de Vries, B., Self, S., Francis, P., Keszthelyi, L., 2001. A gravitational spreading or- Johnson, C.G., Kokelaar, B.P., Iverson, R.M., Logan, M., La Husen, R.G., Gray, J.M.N.T., 2012. igin for the Socompa debris avalanche. J. Volcanol. Geotherm. Res. 105:225–247. Grain-size segregation and levee formation in geophysical mass flows. J. Geophys. https://doi.org/10.1016/S0377-0273(00)00252-3. Res. 117, F01032. https://doi.org/10.01029/02011JF002185. van Wyk de Vries, B., Davies, T., 2015. Landslides, debris avalanches and volcanic gravita- Kokelaar, B.P., Graham, R.L., Gray, J.M.N.T., Vallance, J.W., 2014. Fine-grained linings of lev- tional deformation. In: Sigurdsson, H., et al. (Eds.), Encyclopedia of Volcanoes, 2nd eed channels facilitate runout of granular flows. Earth Planet. Sci. Lett. 385:172–180. edition Elsevier, pp. 665–685. https://doi.org/10.1016/j.epsl.2013.10.043. Luchitta, B.K.A., 1978. Large landslide on Mars. Geol. Soc. Am. Bull. 89, 1601–1609. Malloggi, F., Lanuza, J., Andreotti, B., Clément, E., 2006. Erosion waves: transverse instabil- ities and fingering. Europhys. Lett. 75 (5):825–831. https://doi.org/10.1209/epl/ i2006-10174-10179.

10